Hacker Newsnew | past | comments | ask | show | jobs | submit | Vaibhavk992's commentslogin

Claude 3 Opus is much better at arithmetic than GPT-4. It seems like a combination of tokenizer and well-curated data, the chain of thought formatting closely follows human style.


Bard seems to be a data collection exercise or model alignment in test with the real thing to be released later.


At this point I simply equate Bard with Russias military: much fear for nothing. Still not beaten, but totally useless and laughable either way.

If google intentionally released such an inferior thing while already in panic mode against an existential thread to the company, those deciders must be fired on the spot. Didn’t happen yet, search stopped growing and is _shrinking_ now, and no deciders fired yet, soooo


Hi, Vaibhav here, the creator of the token smuggling attack. They have just banned the variation of this particular prompt, please change the words/smuggling technique and it will work accurately.


I switched a few words and it’s giving me a “something went wrong” error.


Now it's time to hack the guys implementing the fixes. Since they are so fast fixing it they probably don't have time to do much qa.

So design a new jailbreak, advertise it widely, and make sure it's designed in such a way that the fix that the engineers implement creates a much more exploitable and serious vulnerability


So basically be in the same context?


Hooked up ChatGPT with my resume and it now makes a case on why you should hire me for that ML/DL/AI role!


Location: Atlanta, GA, USA

Ask me why am I good fit for role 24x7: https://hirevaibhav.streamlit.app/

Remote: Yes

Willing to relocate: Yes

Technologies: Python, Java, PyTorch, TensorFlow, Keras, Flask, Spark, Hadoop, Pandas, Numpy, Matplotlib, Seaborn, HuggingFace, OpenAI API

Résumé/CV: https://drive.google.com/file/d/1-1FttefUxEo6Rk38J3bmVBszIQF...

Email: vaibhavk [at] gatech [dot] edu

LinkedIn: https://www.linkedin.com/in/vaibhavkumar97/

Experience: MSCS student at Georgia Tech, specializing in ML. Applied science internships at Amazon/Microsoft, published author at EMNLP main conference (full paper), 1 year backend-developer experience.

Looking for full-time ML Engineer/Applied Science/Data science roles starting from Summer 2023.

Extras: Demonstrated the first public jailbreak of LLM behind Bing Chat https://mobile.twitter.com/vaibhavk97/status/162355799717904...


Using virtual functions (masked language modelling and autoregressive prediction) to smuggle banned/sensitive token interaction in a given prompt.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: